Week2/m4 tcp handshake

[Shuhan6017]

Summary

Real TCP handshake for module 4 (Week 2, closes #13).

• send_line/receive_line module-level helpers, with a per-connection
  leftover buffer so partial reads (a line split across two recv()s) and
  batched reads (two lines in one recv()) both work.
• RemoteServer: binds and listens in __init__ so self.port is readable
  immediately (needed for tests on an OS-assigned port), serve_forever()
  loops on accept() and never lets a bad client crash it, and the handler
  does AUTH <token> → OK/ERR bad auth, then REF <branch> → the hash
  read straight off .minigit/refs/heads/<branch> (or - if it doesn't
  exist).
• push/pull now open a real socket, run the same handshake, and print
  the remote hash. All connection failures (refused, dropped, bad auth) are
  mapped to NetworkProtocolError — nothing else leaks out.
• New minigit serve --port <n> --token <t> CLI command.
• Object transfer (WANT/OBJ) is still stubbed for Week 6, per the issue.

Validation

• [X ] scripts/quality-check.sh
• [ X] Manual testing, if applicable:

Checklist

• [ X] The change is focused and easy to review.
• [ X] Tests were added or updated for behavior changes.
• [ X] Documentation was updated for user-facing or setup changes.
• [ X] No secrets, generated artifacts, or local-only files are included.

Notes for Reviewers

I have been a bit late on schedule lately, so this project's code was done by AI. I was able to check and confirm the generated code before commiting. There are some I'm not entirely sure, but I'd still like to push & merge it now. I'll validate it soon, but as for now, since no one is using module 4 and a merge is necessary. ill pr now.